Intraday Liquidity Spillovers in Commodity Futures Markets Thomas A.P. de Boer Master Thesis Wageningen University Agricultural Economics and Rural Policy Group Supervisors: dr.ir. C (Koos) Gardebroek prof.dr.ir. JME (Joost) Pennings Date: September 2019 0 Acknowledgements I want to thank dr. ir. Cornelis Gardebroek for his accurate and comprehensive knowledge of economic models and his ability to explain it with patience, prof. dr. ir. Joost M.E. Pennings for introducing me to the world of agricultural finance and marketing and his expertise in this field, dr. Andrés Trujillo-Barrera for sacrificing early mornings to give his decisive opinion, prof. dr. Philip Garcia for his helpful comments on intraday elements of futures markets, Philippe Debie for his after-hour efforts to generate the limit order book data, Marjolein Verhulst for her critical review on the liquidity measure, Erik Wolters and Koen van Delft from Deloitte Nederland for their outside opinion, my parents for their support, and Marie-Eline Bulten for her unconditional encouragement and optimism. 1 Abstract In this thesis intraday liquidity relations in the soybean crush complex (meal, oil and beans) are examined by studying idiosyncratic and cross-market spillovers of liquidity. Liquidity is a major determinant of derivative pricing, hedging effectiveness and a key driver of co- movements of prices and price-volatilities among markets. A comprehensive multidimensional liquidity measure is derived from the full limited order book (LOB) of the Chicago Mercantile Exchange (CME) for January 2015 to December 2015. A Vector Heterogenous Autoregressive (VHAR) model is adapted to estimate high-resoluted idiosyncratic and cross-market liquidity spillovers in the short-, medium-, and long-run. Results show that liquidity is mostly determined by its own liquidity returns within 30 seconds. Positive cross-market spillovers predominantly occur within 5 minutes and negative cross-market spillovers occur in lags from 5 minutes until one trading day. There is evidence for a so-called ‘flight-to-liquidity’ on a daily time window. Each market provides consistent spillovers to all other markets. During the pre- harvest period the nature of spillovers tends to deviate and the ‘leading’ liquidity role of the soybean market is more pronounced in this period. Furthermore, spillovers differ in nature between regular and extended trading hours, potentially because of different trading strategies. Keywords: Liquidity spillovers, limit order book, futures markets, commodity markets, multivariate HAR model JEL classifications: G13, C22, Q14 2 Table of Contents 1. Introduction .......................................................................................................................... 4 2. Liquidity and its Measurement ........................................................................................... 7 3. Methodology and Data ...................................................................................................... 11 3.1. Liquidity measure .......................................................................................................... 11 3.2. Spillovers analysis using a VHAR model ...................................................................... 12 3.3. Data ............................................................................................................................... 14 4. Results ................................................................................................................................. 19 4.1 Lag structure I ................................................................................................................ 19 4.2. Lag structure II ............................................................................................................. 21 4.3. Regular and Extended Trading Hours .......................................................................... 22 4.4. Seasonality .................................................................................................................... 24 4.5. Robustness checks ......................................................................................................... 24 5. Conclusions and Discussion .............................................................................................. 26 References ............................................................................................................................... 28 3 1. Introduction Co-movements in price returns and their volatility significantly impact pricing in commodity futures markets, and influence decisions for portfolio and risk management. Since commodity futures markets serve both as an asset class for investors and as a risk sharing mechanism for hedgers, common factors that influence pricing are strikingly important in commodity futures markets. The existence of co-movements of commodity prices have been a topic of debate over the last decades. Pindyck & Rotemberg (1990) argue that unrelated raw commodities show co- movement beyond economic fundamental causes. More recently, Ai et al. (2006) present evidence against this excessive commodity price co-movement phenomenon. De Nicola et al. (2016) argue that co-movements among major agricultural, energy, and food commodity are widely present and have been booming over the last years. While scholars do not agree on this topic, these co-movements potentially have great impact. As independence of asset-pricing decreases, shocks to one asset could have serious market wide adverse effects. Since the strong relationship between price volatility and liquidity in futures markets is well-embodied in the literature it is interesting to assess liquidity co-movements (Bessembinder and Seguin, 1993; Pastor and Stambaugh, 2003; Acharya and Pedersen, 2005; Szymanowska et al., 2014). Cross- market liquidity co-movements could potentially explain cross-market price and price volatility correlations, as is discussed by Zhang & Ding (2018) who link co-movement of commodity price returns and their volatilities to liquidity co-movements. A substantial number of studies has focused on co-movements of returns and price volatilities between commodity markets but less studies have assessed liquidity commonalities and spillovers. Chordia et al. (2001) were the first to refer to commonality in liquidity by exploring potential common underlying determinants of microstructure market phenomena. The authors show that liquidity of NYSE stocks correlates with market- and industry wide liquidity, controlling for well-known individual liquidity influencers such as volatility, volume, and price. According to the authors, liquidity commonality indicates that inventory risk and asymmetric information both have an effect on the liquidity of one asset. Hasbrouck and Seppi (2001), Huberman and Halka (2001), and Brockman and Chung (2002) further emphasize the major role of common factors (e.g. liquidity commonality) on the microstructure of markets. Market wide commonalities and potential financial contagion is researched by Rösch and Kaserer (2013), who find significant liquidity commonality in the Xetra electronic market of Deutsche Börse based on a volume-weighted spread measure for liquidity. Not only do the authors find evidence that there is indeed liquidity commonality, they also identify that liquidity commonality increases during market downturns, is large during crisis events, and becomes weaker the further one digs into the Limit Order Book (LOB). The nature of liquidity 4 relations and their magnitudes are especially interesting for agricultural commodity markets, due to potential seasonal production cycle effects and perishability. Furthermore, due to the diverse strategies of traders and their activity, liquidity relations may differ from day to night trading (i.e. Regular Trading Hours (RTH) vs. Extended Trading Hours (ETH)). Mancini et al. (2013) further emphasize liquidity downturns or shocks, as they state that the liquidity on individual foreign exchange rate markets are mostly determined by market wide liquidity shocks. This damages the positive effect of portfolio diversification and implies a relative high risk of liquidity dry-ups in market downturns. The scarce literature on liquidity relations and spillovers mainly focuses on stock markets and equity derivatives. Other markets are less researched, while their characteristics can give interesting insights. For instance, agricultural commodity futures markets are important to study given their unique financial relevance as an asset class and hedge mechanism. Moreover, due to both physical and financial interrelations among various commodity futures markets interesting liquidity dynamics may occur. Our dataset, the soybean complex, gives the opportunity to examine liquidity spillovers in a relative stable environment given the relative balanced relationship between soybeans and the quantity of meal and oil produced (Simon, 1999; Mitchell, 2010). The objective of this study is to examine intraday liquidity relations among futures markets in the soybean complex and to explore the reaction span and persistence of these liquidity spillovers over time. In addition, this study examines differences between liquidity spillovers during regular and extended trading hours and discrepancies due to seasonal effects. Understanding liquidity dynamics can clarify systematic liquidity crises, for example the 2010 Flash Crash (Kirilenko et al., 2017). Furthermore, liquidity plays a substantial